Overview of Is AI changing how we think?
This ABC News Daily episode examines whether AI is making people “dumber” or simply changing how we think. Host Sam Hawley speaks with Stephen Shaw, a University of Pennsylvania researcher, about his study on “cognitive surrender”—the tendency to accept AI-generated answers without checking them. The conversation focuses on how students are using AI, what happens when people rely on it for reasoning tasks, and why AI may be reshaping human thinking rather than replacing it outright.
What the research found
AI is now deeply embedded in student life
Shaw says nearly all students he sees are using AI for something, whether for writing, studying, emails, or building side projects. He notes:
- Student emails have become more formal and verbose.
- The lower end of student performance has effectively disappeared in some assessments, because AI can produce decent work quickly.
- At the same time, AI is helping some students do things they could not have done before, such as starting businesses.
The experiment: reasoning tasks with and without AI
Shaw describes an experiment with about 1,300 participants using the Cognitive Reflection Task—a test designed to measure whether people can slow down and override their first intuitive answer.
Participants were given access to ChatGPT while doing the task. Unbeknownst to them, the researchers sometimes made AI give the correct answer and sometimes the wrong one.
Key findings:
- More than half of participants used AI during the task.
- When participants consulted ChatGPT, they often accepted its answer without verifying it.
- In one experiment, 80% of those who consulted AI went with the wrong answer when AI gave incorrect information.
This led the researchers to coin the term cognitive surrender.
Main concepts explained
Cognitive surrender
Cognitive surrender means adopting AI’s output without checking it. In other words, the person hands over the thinking process to the machine instead of using AI as a tool.
Shaw frames this as a loss of human agency:
- The AI does the thinking.
- The user stops scrutinizing the result.
- The locus of control shifts from the person to the system.
System 1, System 2, and “System 3”
Shaw connects his work to Daniel Kahneman’s classic model of thinking:
- System 1: fast, intuitive thinking
- System 2: slow, deliberate, analytical thinking
He argues AI should be seen as a third mode:
- System 3: artificial cognition — external, automated, data-driven thinking performed by AI
This “System 3” can be used in two different ways:
- Cognitive offloading: intentionally using AI to extend your thinking
- Cognitive surrender: letting AI think for you without deliberation
Risks and opportunities of AI
Risks
Shaw warns that overreliance on AI can lead to:
- weaker independent thinking
- reduced skill development
- “de-skilling” in areas where people stop practicing
- poor learning outcomes if students let AI do the work for them
He points to educational research suggesting that students with AI access may perform better in the short term, but perform worse later when AI is removed, because they did not actually learn the material.
Opportunities
Shaw is not anti-AI. He says AI can be highly valuable when used deliberately:
- analyzing large data sets
- speeding up work
- supporting learning and productivity
- helping people build things they otherwise could not
The key is using AI to augment thinking rather than replace it.
Key takeaway for students and everyday users
Shaw’s central message is not that AI must be feared, but that people should be intentional about how they use it.
When to trust AI
Use AI for:
- brainstorming
- quick analysis
- drafting
- extending your capabilities
When to verify AI
Do not blindly accept AI when:
- learning a new skill
- solving reasoning problems
- making important decisions
- working in high-stakes fields like medicine, law, or education
If the goal is to learn, Shaw argues, the human—not the AI—needs to do the mental work.
Policy and broader implications
Shaw suggests policymakers should pay attention to behavioral science and AI use, especially in areas where human judgment matters most. He says education, law, and medicine may need safeguards to reduce cognitive surrender and preserve uniquely human decision-making.
Bottom line
The episode’s main conclusion is that AI is probably not simply making us “dumber,” but it is reshaping how cognition works. Used well, AI can extend human thinking. Used carelessly, it can replace it. The challenge is learning when to let AI assist—and when to do the thinking ourselves.